Natura-1 Trains Two 20-DoF Hands From Human Video

Loading video
Loading videoNature Will trained Natura-1 on one million hours of first-person video, using a 50B-parameter mixture-of-experts model to turn language and camera images into dual-arm and 20-DoF hand actions within a roughly 100 ms loop.
Category: humanoid
Author: @LeoKharon
Date: 2026-11-10T00:00:00
Duration: 81.199s





